pygpt-net 2.6.31__py3-none-any.whl → 2.6.32__py3-none-any.whl

pygpt_net/CHANGELOG.txt

@@ -1,3 +1,10 @@
+2.6.32 (2025-09-02)
+
+- Added video generation and support for Google Veo 3 models.
+- Introduced new predefined models: veo-3.0-generate-preview and veo-3.0-fast-generate-preview.
+- Integrated DuckDuckGo as a search provider in the WebSearch plugin.
+- Added "Loop" mode to Realtime + audio mode for automatic turn handling and continuous conversation without manually enabling the microphone.
+
 2.6.31 (2025-09-01)
 
 - Chat with Audio mode renamed to Realtime + audio.

pygpt_net/__init__.py

@@ -6,15 +6,15 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.09.01 00:00:00                  #
+# Updated Date: 2025.09.02 00:00:00                  #
 # ================================================== #
 
 __author__ = "Marcin Szczygliński"
 __copyright__ = "Copyright 2025, Marcin Szczygliński"
 __credits__ = ["Marcin Szczygliński"]
 __license__ = "MIT"
-__version__ = "2.6.31"
-__build__ = "2025-09-01"
+__version__ = "2.6.32"
+__build__ = "2025-09-02"
 __maintainer__ = "Marcin Szczygliński"
 __github__ = "https://github.com/szczyglis-dev/py-gpt"
 __report__ = "https://github.com/szczyglis-dev/py-gpt/issues"

pygpt_net/app.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.27 20:00:00                  #
+# Updated Date: 2025.09.02 01:00:00                  #
 # ================================================== #
 
 import os
@@ -182,6 +182,7 @@ from pygpt_net.provider.audio_output.eleven_labs import ElevenLabsTextToSpeech
 # web search engine providers
 from pygpt_net.provider.web.google_custom_search import GoogleCustomSearch
 from pygpt_net.provider.web.microsoft_bing import MicrosoftBingSearch
+from pygpt_net.provider.web.duckduck_search import DuckDuckGoSearch
 
 # tools
 from pygpt_net.tools.indexer import IndexerTool
@@ -342,6 +343,7 @@ def run(**kwargs):
     # register web providers
     launcher.add_web(GoogleCustomSearch())
     launcher.add_web(MicrosoftBingSearch())
+    launcher.add_web(DuckDuckGoSearch())
 
     # register custom web providers
     providers = kwargs.get('web', None)

pygpt_net/app_core.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.28 09:00:00                  #
+# Updated Date: 2025.09.01 23:00:00                  #
 # ================================================== #
 
 from pygpt_net.config import Config
@@ -43,6 +43,7 @@ from pygpt_net.core.tabs import Tabs
 from pygpt_net.core.text import Text
 from pygpt_net.core.tokens import Tokens
 from pygpt_net.core.updater import Updater
+from pygpt_net.core.video import Video
 from pygpt_net.core.vision import Vision
 from pygpt_net.core.web import Web
 
@@ -92,6 +93,7 @@ class Core:
         self.text = Text(window)
         self.tokens = Tokens(window)
         self.updater = Updater(window)
+        self.video = Video(window)
         self.vision = Vision(window)
         self.web = Web(window)
 

pygpt_net/config.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.18 01:00:00                  #
+# Updated Date: 2025.09.01 23:00:00                  #
 # ================================================== #
 
 import copy
@@ -66,6 +66,8 @@ class Config:
             "presets": "presets",
             "upload": "upload",
             "tmp": "tmp",
+            "video": "video",
+            "music": "music",
         }
         self._app_path = None
         self._version_cache = version if version else None

pygpt_net/controller/__init__.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.30 06:00:00                  #
+# Updated Date: 2025.09.01 23:00:00                  #
 # ================================================== #
 
 from pygpt_net.controller.access import Access
@@ -29,6 +29,7 @@ from pygpt_net.controller.kernel import Kernel
 from pygpt_net.controller.lang import Lang
 from pygpt_net.controller.launcher import Launcher
 from pygpt_net.controller.layout import Layout
+from pygpt_net.controller.media import Media
 from pygpt_net.controller.mode import Mode
 from pygpt_net.controller.model import Model
 from pygpt_net.controller.notepad import Notepad
@@ -71,6 +72,7 @@ class Controller:
         self.lang = Lang(window)
         self.launcher = Launcher(window)
         self.layout = Layout(window)
+        self.media = Media(window)
         self.mode = Mode(window)
         self.model = Model(window)
         self.notepad = Notepad(window)
@@ -111,6 +113,7 @@ class Controller:
         self.camera.setup_ui()
         self.access.setup()
         self.realtime.setup()
+        self.media.setup()
 
     def post_setup(self):
         """Post-setup, after plugins are loaded"""
@@ -169,6 +172,7 @@ class Controller:
         self.lang.reload()
         self.debug.reload()
         self.chat.reload()
+        self.media.reload()
         self.window.tools.on_reload()
         self.access.reload()
         self.tools.reload()

pygpt_net/controller/audio/audio.py

@@ -40,12 +40,19 @@ class Audio:
     def setup(self):
         """Setup controller"""
         self.update()
+
+        # continuous input (notepad)
         if self.window.core.config.get("audio.input.continuous", False):
             self.window.ui.plugin_addon['audio.input.btn'].continuous.setChecked(True)
 
+        # auto turn (VAD)
         if self.window.core.config.get("audio.input.auto_turn", False):
             self.window.ui.nodes['audio.auto_turn'].box.setChecked(True)
 
+        # loop recording
+        if self.window.core.config.get("audio.input.loop", False):
+            self.window.ui.nodes['audio.loop'].box.setChecked(True)
+
     def execute_input_stop(self):
         """Execute input stop (from UI)"""
         self.window.dispatch(Event(Event.AUDIO_INPUT_RECORD_TOGGLE, {
@@ -67,6 +74,12 @@ class Audio:
         self.window.core.config.set("audio.input.auto_turn", value)
         self.window.core.config.save()
 
+    def toggle_loop(self):
+        """Toggle loop recording setting"""
+        value = self.window.ui.nodes['audio.loop'].box.isChecked()
+        self.window.core.config.set("audio.input.loop", value)
+        self.window.core.config.save()
+
     def toggle_input(
             self,
             state: bool,

pygpt_net/controller/chat/common.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.27 07:00:00                  #
+# Updated Date: 2025.09.01 23:00:00                  #
 # ================================================== #
 
 import os
@@ -100,37 +100,6 @@ class Common:
         event = RenderEvent(RenderEvent.ON_SWITCH)
         self.window.dispatch(event)  # switch renderer if needed
 
-        # edit icons
-        """
-        if self.window.core.config.has('ctx.edit_icons'):
-            self.window.ui.nodes['output.edit'].setChecked(self.window.core.config.get('ctx.edit_icons'))
-            data = {
-                "initialized": self.initialized,
-            }
-            if self.window.core.config.get('ctx.edit_icons'):
-                event = RenderEvent(RenderEvent.ON_EDIT_ENABLE, data)
-            else:
-                event = RenderEvent(RenderEvent.ON_EDIT_DISABLE, data)
-            self.window.dispatch(event)
-        """
-
-        # images generation
-        if self.window.core.config.get('img_raw'):
-            self.window.ui.config['global']['img_raw'].setChecked(True)
-        else:
-            self.window.ui.config['global']['img_raw'].setChecked(False)
-
-        # image resolution
-        resolution = self.window.core.config.get('img_resolution', '1024x1024')
-        self.window.controller.config.apply_value(
-            parent_id="global",
-            key="img_resolution",
-            option=self.window.core.image.get_resolution_option(),
-            value=resolution,
-        )
-        if not self.initialized:
-            self.window.ui.add_hook("update.global.img_resolution", self.hook_update)
-
         # set focus to input
         self.window.ui.nodes['input'].setFocus()
         self.initialized = True
@@ -404,29 +373,6 @@ class Common:
             event = RenderEvent(RenderEvent.ON_TS_DISABLE, data)
         self.window.dispatch(event)
 
-    def toggle_raw(self, value: bool):
-        """
-        Toggle raw (plain) output
-
-        :param value: value of the checkbox
-        """
-        self.window.core.config.set('render.plain', value)
-        self.window.core.config.save()
-
-        # update checkbox in settings dialog
-        self.window.controller.config.checkbox.apply(
-            'config',
-            'render.plain',
-            {
-                'value': value
-            },
-        )
-        event = RenderEvent(RenderEvent.ON_SWITCH)
-        self.window.dispatch(event)
-
-        # restore previous font size
-        self.window.controller.ui.update_font_size()
-
     def toggle_edit_icons(self, value: bool):
         """
         Toggle edit icons
@@ -444,39 +390,28 @@ class Common:
             event = RenderEvent(RenderEvent.ON_EDIT_DISABLE, data)
         self.window.dispatch(event)
 
-    def img_enable_raw(self):
-        """Enable help for images"""
-        self.window.core.config.set('img_raw', True)
-        self.window.core.config.save()
-
-    def img_disable_raw(self):
-        """Disable help for images"""
-        self.window.core.config.set('img_raw', False)
-        self.window.core.config.save()
-
-    def img_toggle_raw(self, state: bool):
+    def toggle_raw(self, value: bool):
         """
-        Toggle help for images
+        Toggle raw (plain) output
 
-        :param state: state of checkbox
+        :param value: value of the checkbox
         """
-        if not state:
-            self.img_disable_raw()
-        else:
-            self.img_enable_raw()
+        self.window.core.config.set('render.plain', value)
+        self.window.core.config.save()
 
-    def hook_update(self, key: str, value: Any, caller, *args, **kwargs):
-        """
-        Hook for updating image resolution
+        # update checkbox in settings dialog
+        self.window.controller.config.checkbox.apply(
+            'config',
+            'render.plain',
+            {
+                'value': value
+            },
+        )
+        event = RenderEvent(RenderEvent.ON_SWITCH)
+        self.window.dispatch(event)
 
-        :param key: config key
-        :param value: new value
-        :param caller: caller object
-        """
-        if key == "img_resolution":
-            if not value:
-                return
-            self.window.core.config.set('img_resolution', value)
+        # restore previous font size
+        self.window.controller.ui.update_font_size()
 
     def save_text(
             self,

pygpt_net/controller/lang/custom.py

@@ -55,8 +55,8 @@ class Custom:
         self.window.ui.config['preset'][MODE_CHAT].box.setText(trans("preset.chat"))
         self.window.ui.config['preset'][MODE_COMPLETION].box.setText(trans("preset.completion"))
         self.window.ui.config['preset'][MODE_IMAGE].box.setText(trans("preset.img"))
-        self.window.ui.config['preset'][MODE_VISION].box.setText(trans("preset.vision"))
-        #self.window.ui.config['preset'][MODE_LANGCHAIN].box.setText(trans("preset.langchain"))
+        # self.window.ui.config['preset'][MODE_VISION].box.setText(trans("preset.vision"))
+        # self.window.ui.config['preset'][MODE_LANGCHAIN].box.setText(trans("preset.langchain"))
         self.window.ui.config['preset'][MODE_LLAMA_INDEX].box.setText(trans("preset.llama_index"))
         self.window.ui.config['preset'][MODE_AGENT].box.setText(trans("preset.agent"))
         self.window.ui.config['preset'][MODE_AGENT_LLAMA].box.setText(trans("preset.agent_llama"))

pygpt_net/controller/media/__init__.py

@@ -0,0 +1,12 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+# ================================================== #
+# This file is a part of PYGPT package               #
+# Website: https://pygpt.net                         #
+# GitHub:  https://github.com/szczyglis-dev/py-gpt   #
+# MIT License                                        #
+# Created By  : Marcin Szczygliński                  #
+# Updated Date: 2025.09.01 23:00:00                  #
+# ================================================== #
+
+from .media import Media

pygpt_net/controller/media/media.py

@@ -0,0 +1,115 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+# ================================================== #
+# This file is a part of PYGPT package               #
+# Website: https://pygpt.net                         #
+# GitHub:  https://github.com/szczyglis-dev/py-gpt   #
+# MIT License                                        #
+# Created By  : Marcin Szczygliński                  #
+# Updated Date: 2025.09.01 23:00:00                  #
+# ================================================== #
+
+from typing import Any
+
+
+class Media:
+    def __init__(self, window=None):
+        """
+        Media (video, image, music) controller
+
+        :param window: Window instance
+        """
+        self.window = window
+        self.initialized = False
+
+    def setup(self):
+        """Setup UI"""
+        # raw mode for images/video
+        if self.window.core.config.get('img_raw'):
+            self.window.ui.config['global']['img_raw'].setChecked(True)
+        else:
+            self.window.ui.config['global']['img_raw'].setChecked(False)
+
+        # image: resolution
+        resolution = self.window.core.config.get('img_resolution', '1024x1024')
+        self.window.controller.config.apply_value(
+            parent_id="global",
+            key="img_resolution",
+            option=self.window.core.image.get_resolution_option(),
+            value=resolution,
+        )
+
+        # video: aspect ratio
+        aspect_ratio = self.window.core.config.get('video.aspect_ratio', '16:9')
+        self.window.controller.config.apply_value(
+            parent_id="global",
+            key="video.aspect_ratio",
+            option=self.window.core.video.get_aspect_ratio_option(),
+            value=aspect_ratio,
+        )
+
+        # -- add hooks --
+        if not self.initialized:
+            self.window.ui.add_hook("update.global.img_resolution", self.hook_update)
+            self.window.ui.add_hook("update.global.video.aspect_ratio", self.hook_update)
+
+    def reload(self):
+        """Reload UI"""
+        self.setup()
+
+    def hook_update(self, key: str, value: Any, caller, *args, **kwargs):
+        """
+        Hook for updating media options
+
+        :param key: config key
+        :param value: new value
+        :param caller: caller object
+        """
+        if key == "img_resolution":
+            if not value:
+                return
+            self.window.core.config.set('img_resolution', value)
+        elif key == "video.aspect_ratio":
+            if not value:
+                return
+            self.window.core.config.set('video.aspect_ratio', value)
+
+    def enable_raw(self):
+        """Enable prompt enhancement for images"""
+        self.window.core.config.set('img_raw', True)
+        self.window.core.config.save()
+
+    def disable_raw(self):
+        """Disable prompt enhancement for images"""
+        self.window.core.config.set('img_raw', False)
+        self.window.core.config.save()
+
+    def toggle_raw(self):
+        """Save prompt enhancement option for images"""
+        state = self.window.ui.config['global']['img_raw'].isChecked()
+        if not state:
+            self.disable_raw()
+        else:
+            self.enable_raw()
+
+    def is_image_model(self) -> bool:
+        """
+        Check if the model is an image generation model
+
+        :return: True if the model is an image generation model
+        """
+        current = self.window.core.config.get("model")
+        model_data = self.window.core.models.get(current)
+        if model_data:
+            return model_data.is_image_output()
+
+    def is_video_model(self) -> bool:
+        """
+        Check if the model is a video generation model
+
+        :return: True if the model is a video generation model
+        """
+        current = self.window.core.config.get("model")
+        model_data = self.window.core.models.get(current)
+        if model_data:
+            return model_data.is_video_output()

pygpt_net/controller/realtime/realtime.py

@@ -11,7 +11,14 @@
 
 from PySide6.QtCore import Slot, QTimer
 
-from pygpt_net.core.events import RealtimeEvent, RenderEvent, BaseEvent, AppEvent, KernelEvent
+from pygpt_net.core.events import (
+    RealtimeEvent,
+    RenderEvent,
+    BaseEvent,
+    AppEvent,
+    KernelEvent,
+    Event,
+)
 from pygpt_net.core.realtime.worker import RealtimeSignals
 from pygpt_net.core.types import MODE_AUDIO
 from pygpt_net.utils import trans
@@ -122,10 +129,12 @@ class Realtime:
                     "begin": False,
                 }))
 
-        # audio end: stop audio playback
+        # audio end: on stop audio playback
         elif event.name == RealtimeEvent.RT_OUTPUT_AUDIO_END:
             self.set_idle()
             self.window.controller.chat.common.unlock_input()
+            if self.is_loop():
+                QTimer.singleShot(500, lambda: self.next_turn())  # wait a bit before next turn
 
         # end of turn: finalize the response
         elif event.name == RealtimeEvent.RT_OUTPUT_TURN_END:
@@ -163,6 +172,22 @@ class Realtime:
         elif event.name == AppEvent.CTX_SELECTED:
             QTimer.singleShot(0, lambda: self.reset())
 
+    def next_turn(self):
+        """Start next turn in loop mode (if enabled)"""
+        self.window.dispatch(Event(Event.AUDIO_INPUT_RECORD_TOGGLE))
+        if self.window.controller.audio.is_recording():
+            QTimer.singleShot(100, lambda: self.window.update_status(trans("speech.listening")))
+
+    def is_loop(self) -> bool:
+        """
+        Check if loop recording is enabled
+
+        :return: True if loop recording is enabled, False otherwise
+        """
+        if self.window.controller.kernel.stopped():
+            return False
+        return self.window.core.config.get("audio.input.loop", False)
+
     @Slot(object)
     def handle_response(self, event: RealtimeEvent):
         """

pygpt_net/controller/ui/mode.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.15 23:00:00                  #
+# Updated Date: 2025.09.01 23:00:00                  #
 # ================================================== #
 
 from pygpt_net.core.types import (
@@ -60,8 +60,10 @@ class Mode:
 
         if not is_audio:
             self.window.ui.nodes['audio.auto_turn'].setVisible(False)
+            self.window.ui.nodes["audio.loop"].setVisible(False)
         else:
             self.window.ui.nodes['audio.auto_turn'].setVisible(True)
+            self.window.ui.nodes["audio.loop"].setVisible(True)
 
         if not is_assistant:
             ui_nodes['presets.widget'].setVisible(True)
@@ -138,9 +140,21 @@ class Mode:
             ui_tabs['preset.editor.extra'].setTabText(0, trans("preset.prompt"))
 
         if is_image:
-            ui_nodes['dalle.options'].setVisible(True)
+            ui_nodes['media.raw'].setVisible(True)
+            if ctrl.media.is_video_model():
+                ui_nodes['video.options'].setVisible(True)
+                ui_nodes['dalle.options'].setVisible(False)
+            elif ctrl.media.is_image_model():
+                ui_nodes['dalle.options'].setVisible(True)
+                ui_nodes['video.options'].setVisible(False)
+            else:
+                ui_nodes['media.raw'].setVisible(False)
+                ui_nodes['dalle.options'].setVisible(False)
+                ui_nodes['video.options'].setVisible(False)
         else:
+            ui_nodes['media.raw'].setVisible(False)
             ui_nodes['dalle.options'].setVisible(False)
+            ui_nodes['video.options'].setVisible(False)
 
         if is_agent:
             ui_nodes['agent.options'].setVisible(True)

pygpt_net/core/audio/backend/pyaudio/realtime.py

@@ -1,14 +1,3 @@
-#!/usr/bin/env python3
-# -*- coding: utf-8 -*-
-# ================================================== #
-# This file is a part of PYGPT package               #
-# Website: https://pygpt.net                         #
-# GitHub:  https://github.com/szczyglis-dev/py-gpt   #
-# MIT License                                        #
-# Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.31 23:00:00                  #
-# ================================================== #
-
 import threading
 from typing import Optional
 
@@ -53,6 +42,9 @@ class RealtimeSessionPyAudio(QObject):
         self._final = False
         self._tail_ms = 60  # add a small silence tail to avoid clicks
 
+        # one-shot guard to avoid double stop and duplicate callbacks
+        self._stopping = False
+
         # volume metering
         self._volume_emitter = volume_emitter
         self._vol_buffer = bytearray()
@@ -78,6 +70,13 @@ class RealtimeSessionPyAudio(QObject):
         except Exception:
             pass
 
+        # finished-state watchdog: guarantees stop()+on_stopped once playback is truly done
+        self._finish_timer = QTimer(self)
+        self._finish_timer.setTimerType(Qt.PreciseTimer)
+        self._finish_timer.setInterval(15)  # fast but lightweight watchdog
+        self._finish_timer.timeout.connect(self._check_finished)
+        self._finish_timer.start()
+
         # stop callback (set by backend)
         self.on_stopped = None
 
@@ -124,15 +123,28 @@ class RealtimeSessionPyAudio(QObject):
         self._final = True
 
     def stop(self) -> None:
-        """Stop playback and free resources."""
+        """Stop playback and free resources. Idempotent."""
+        # ensure this executes only once even if called from multiple paths
+        if self._stopping:
+            return
+        self._stopping = True
+
+        # stop timers first to prevent re-entry
+        try:
+            if self._finish_timer:
+                self._finish_timer.stop()
+        except Exception:
+            pass
         try:
             if self._vol_timer:
                 self._vol_timer.stop()
         except Exception:
             pass
+
+        # gracefully stop PortAudio stream and close/terminate
         try:
             if self._stream and self._stream.is_active():
-                self._stream.stop_stream()
+                self._stream.stop_stream()  # drains queued audio per PortAudio docs
         except Exception:
             pass
         try:
@@ -197,11 +209,36 @@ class RealtimeSessionPyAudio(QObject):
 
         # auto-finish: when final and nothing more to play, complete and stop()
         if self._final and self._buffer_empty():
-            QTimer.singleShot(0, self.stop)  # stop on the GUI thread
+            # Return paComplete and request stop on the GUI thread.
+            # PaComplete deactivates the stream after the last callback buffer is played.
+            QTimer.singleShot(0, self.stop)
             return out, pyaudio.paComplete
 
         return out, pyaudio.paContinue
 
+    def _check_finished(self) -> None:
+        """
+        Watchdog that runs on the Qt thread to guarantee a single, reliable stop().
+        Triggers when PortAudio deactivates the stream, or when the buffer is fully
+        drained after mark_final().
+        """
+        if self._stopping:
+            return
+
+        # If underlying PA stream is no longer active, we are done.
+        try:
+            if self._stream is not None and not self._stream.is_active():
+                self.stop()
+                return
+        except Exception:
+            # If querying state fails, assume the stream is done and stop.
+            self.stop()
+            return
+
+        # If we've been marked final and our buffer is empty, finalize proactively.
+        if self._final and self._buffer_empty():
+            self.stop()
+
     def _buffer_empty(self) -> bool:
         """
         Check if internal buffer is empty.

pygpt_net/core/audio/output.py

@@ -43,8 +43,9 @@ class Output:
         return self.backends[backend]
 
     def setup(self):
-        """Setup audio output backend"""
-        pass
+        """Setup audio input backend"""
+        for b in self.backends.values():
+            b.set_rt_signals(self.window.controller.realtime.signals)
 
     def play(
             self,